Professional Documents
Culture Documents
& 2017 The Author(s) Published by the Royal Society. All rights reserved.
for sleep and activity at particular times during a 24 h period) [3]. patterns that correspond with wakefulness or light sleep. 2
Individuals in the tails of the distribution are colloquially known Alternatively, if all individuals in the study tend to be asleep
rspb.royalsocietypublishing.org
as morning ‘larks’ and night ‘owls,’ reflecting that they either go simultaneously for long intervals of time, the sentinel hypo-
to sleep early and wake early, or go to sleep late and wake late. thesis would be rejected. To test this general prediction, we
Chronotype variation is found across animals; examples include developed a simulation model to produce a null distribution
the existence of stable, species-typical chronotypes in birds [4,5] of wakefulness overlap that is tailored to our sampling
and rodents [6], and the chronotype of the diurnal degus regime and data. Empirical observations of more extreme vari-
(Octodon degus) is described as similar to humans [7]. ation in wakefulness overlap would support the sentinel
In Western populations, the chronotype distribution is nor- hypothesis. We further used this model to investigate the fac-
mally distributed, sex-specific (with males biased towards tors that influence sleep synchrony, focusing on variation in
being late chronotype) and characterized by systematic chronotype and group size. Finally, to assess whether demo-
changes to chronotype across the lifespan (with older individ- graphic and other factors maintain chronotype variation, we
rspb.royalsocietypublishing.org
24 h period throughout the study (n ¼ 20 days), we performed
material for the English version of the administered survey). an epoch-by-epoch analysis, from the phase at which the first
subject fell asleep to the phase at which the last subject awoke.
Sleep epochs were determined by the CamNtech MotionWare
software after the sleep onset and sleep offset periods were desig-
(b) Protocol nated manually by an experienced scorer (D.R.S.), guided by the
Participants were verbally administered a sleep survey in Swahili, event markers used by study participants. We generated descrip-
their second language, at the beginning of the study period tive statistics reporting the mean wake-ratio ( proportion of
(20 January–11 February 2016) to screen for healthy sleep and subjects that are awake at any given night-time epoch) and
ascertain general information on subjective sleep quality and mean of the median of subjects awake at any given night-time
threat perception (see electronic supplemental material). During epoch. We also generated the effective group sleep time, which is
rspb.royalsocietypublishing.org
person falls asleep and the last person wakes up. The individual average
for this group is measured as sleep onset and sleep end; note that the
3. Results additional time between onset and end is the individual averaged wake
after sleep onset (WASO) time of 2.4 h.
(a) Sleep sites and threat perception
The camp consisted of two subgroups of grass-huts (total effective individual
n ¼ 22). The distance between the two groups was 127 m, parameter group sleep averaged sleep
measured to the centre of each. The average distance
between huts was 12.7 m, as measured to the closest adja- sleep start 20.34 + 31 min 22.13 + 36 min
cent hut. The average hut size was 1.89 m (s.d. ¼ 0.42) tall, sleep end 07.48 + 20 min 06.55 + 28 min
rspb.royalsocietypublishing.org
activity
1500
1000
500
0
midnight 6 AM noon 6 PM midnight
time
Figure 1. Functional linear modelling (FLM) analysis. Individual 24 h averaged activity (counts per minute) generated from FLM is compared among all individual
400
measure
04.00
200
02.00
100
0
20 30 40 50 60
age
circadian phase measure of Hadza individuals
Figure 2. Circadian phase measure (CPM) in Hadza hunter – gatherers. CPM Figure 3. Plotted slope for the fixed coefficient of ‘age.’ Circadian phase
is the midpoint clock time between sleep onset and awakening. Here, the measure (CPM) was regressed on age, with circadian phase generated by cal-
average CPM is illustrated for each individual with a beanplot, where the culating each individual’s nightly number of minutes asleep from after
horizontal white lines show the individual night-time CPM observations midnight. The model showed that sex, number of co-sleeping children, nur-
and the black shows the distribution of the observed data; shape of the bean- sing status and the study day (correcting for weather) did not significantly
plot represents density. ANOVA revealed that the Hadza significantly differ in influence CPM, but age did.
CPM on the individual level (see §3c). Additionally, the more intra-individual
variation, the greater the spread from the red line; that is, the longer and of epochs in which no individuals were scored as awake
thinner the bean plots, the more variation is seen on the intra-individual (n ¼ 18) fell more clearly within the simulated distribution
level. Note that the tails of the CPM distributions overlap several night- (mean ¼ 12.4, 95% of distribution was from 0 to 38, with the
time hours; in other words, variability in sleep timing is distributed through- observed value marking the 90th percentile).
out the night-time period. These results further raise the question of how chronotype
variation is maintained, which we tested cross-sectionally
in a linear mixed effects model. We found that variation in
individual’s activity could promote multiple arousals as a type chronotype (CPM) covaries with age, but not with other
of cascade effect, which could sustain a pool of awake individ- fixed effects that include sex, co-sleeping, nursing status
uals, even if those awake tend to return to sleep. and study day. That is, at older ages, individuals tend to exhi-
We again used our simulation model of independently bit a more ‘lark’ type chronotype (age: b + s.e. ¼ 20.23 +
sleeping individuals to investigate these possibilities, but this 0.09, p ¼ 0.01, CI ¼ 20.401, 20.056; figure 3). Thus, variation
time including chronotype variation across individuals by in age helps to generate variation in chronotype, which then
simulating individual variation in inferred sleep onset and facilitates sentinel-like behaviour.
offset based on the nightly data. We expected that observed We also used the model to investigate the effects of group
group wakefulness measures would fall outside of the simu- size, aiming to assess the relative effects of group size and
lated distributions if active asynchrony is needed to achieve chronotype variation on night-time safety in a more general
the observed level of wakefulness, and within the simulated modelling framework that is not strictly tied to this population
distribution if chronotype variation is sufficient to achieve and sampling regime. The model predicted that the mean
the group wakefulness we observed. We found support for number of individuals awake scales linearly with group size
the effect of chronotype variation on maintaining sentinel- (figure 4a), and that the proportion of the time all individuals
like behaviour: the observed mean of 8.65 individuals scored are asleep drops asymptotically to zero with increasing
as awake was well within the distribution of simulated group size (figure 4b). Without variation in chronotypes, the
values from the model that included chronotype (mean ¼ model revealed that group size would have to be substantially
8.94, 95% of distribution was from 7.73 to 10.2, and 32.3% of larger to reach a similar level of safety as found with variation
simulated values were below 8.65). Similarly, the number in chronotype (figure 4).
(a) actigraphy does not necessarily mean that individuals scored 6
20 as ‘awake’ are truly awake and vigilant, as compared to studies
rspb.royalsocietypublishing.org
in other mammals where more direct observations of sentinel
mean number awake
15 behaviours are possible. Sentinel behaviour has often been
inferred by observation and anecdotal evidence, but rarely
10 has coordinated vigilance been systematically measured in
animal species; thus, in attempts to quantify sentinel behaviour,
5 it has been defined as the distribution of bouts of wakefulness
across time that are more evenly distributed than expected by
0 chance [42]. In meerkats, for example, sentinel behaviour
has been measured as the proportion of time individuals are
10 20 30 40 bipedal with heads raised [42].
rspb.royalsocietypublishing.org
chronotype variation [50,51]. Variability in chronotype could has since been supported in phylogenetic analyses showing
be further enhanced through sexual selection, perhaps as indi- that mammalian species that experience a greater risk of preda-
cated by variation in reproductive success among chronotypes tion at the sleep site exhibited less REM sleep [56,57] and that
in Western populations [52]. Alternatively, greater flexibility in humans have both shorter than expected sleep and a higher
sleep patterns could emerge simply from relaxed selection on a than expected proportion of REM sleep [58]. Throughout
specific chronotype in our evolutionary past. human evolution, the transition to relatively deeper, higher
Our findings also have implications for evolutionary per- quality sleep would have conferred cognitive benefits such as
spectives on sleep disorders. Disorders of circadian rhythm the consolidation of memory [59], processing of emotion lead-
are defined by a mismatch between the desired sleep demands ing to better social intelligence [60], and increasing creativity
of the social environment and that of an individual’s natural and innovation [61]. The success of Homo sapiens has been
References
1. Snyder F. 1966 Toward an evolutionary theory of 7. Labyak SE, Lee TM, Goel N. 1997 Rhythm 12. Randler C, Schaal S. 2010 Morningness –
dreaming. Am. J. Psychiatry 123, 121 –136. (doi:10. chronotypes in a diurnal rodent, Octodon degus. eveningness, habitual sleep–wake variables and
1176/ajp.123.2.121) Am. J. Physiol 273, R1058– R1066. cortisol level. Biol. Psychol. 85, 14 –18. (doi:10.
2. Beauchamp G. 2015 Animal vigilance: monitoring 8. Roenneberg T, Kuehnle T, Pramstaller PP, Ricken J, 1016/j.biopsycho.2010.04.006)
predators and competitors. Quebec, Canada: Havel M, Guth A, Merrow M. 2004 A marker for the 13. Randler C, Ebenhöh N, Fischer A, Höchel S, Schroff
Academic Press. end of adolescence. Curr. Biol. 14, R1038–R1039. C, Stoll JC, Vollmer C. 2012 Chronotype but not
3. Randler C. 2014 Sleep, sleep timing and chronotype (doi:10.1016/j.cub.2004.11.039) sleep length is related to salivary testosterone in
in animal behaviour. Anim. Behav. 94, 161–166. 9. Baehr EK, Revelle W, Eastman CI. 2000 Individual young adult men. Psychoneuroendocrinology 37,
(doi:10.1016/j.anbehav.2014.05.001) differences in the phase and amplitude of the 1740– 1744. (doi:10.1016/j.psyneuen.2012.02.008)
4. Von Aschoff J, Wever R. 1962 Spontanperiodik des human circadian temperature rhythm: with an 14. Lopez-Minguez J, Ordoñana JR, Sánchez-Romera JF,
menschen bei ausschluss aller zeitgeber. emphasis on morningness–eveningness. J. Sleep Madrid JA, Garaulet M. 2015 Circadian system
Naturwissenschaften 49, 337 –342. (doi:10.1007/ Res. 9, 117–127. (doi:10.1046/j.1365-2869.2000. heritability as assessed by wrist temperature: a twin
BF01185109) 00196.x) study. Chronobiol. Int. 32, 71 –80. (doi:10.3109/
5. Steinmeyer C, Schielzeth H, Mueller JC, 10. Sarabia J, Rol M, Mendiola P, Madrid J. 2008 07420528.2014.955186)
Kempenaers B. 2010 Variation in sleep behaviour Circadian rhythm of wrist temperature in normal- 15. Viola AU, Archer SN, James LM, Groeger JA, Lo JCY,
in free-living blue tits, Cyanistes caeruleus: effects of living subjects: a candidate of new index of the Skene DJ, von Schantz M, Dijk DJ. 2007 PER3
sex, age and environment. Anim. Behav. 80, 853– circadian system. Physiol. Behav. 95, 570 –580. polymorphism predicts sleep structure and waking
864. (doi:10.1016/j.anbehav.2010.08.005) (doi:10.1016/j.physbeh.2008.08.005) performance. Curr. Biol. 17, 613 –618. (doi:10.
6. Wicht H, Korf HW, Ackermann H, Ekhart D, Fischer 11. Burgess HJ, Fogg LF. 2008 Individual differences in 1016/j.cub.2007.01.073)
C, Pfeffer M. 2014 Chronotypes and rhythm stability the amount and timing of salivary melatonin 16. Patke A, Murphy PJ, Onat OE, Krieger AC, Özçelik T,
in mice. Chronobiol. Int. 31, 27 –36. (doi:10.3109/ secretion. PLoS ONE 3, e3055. (doi:10.1371/journal. Campbell SS, Young MW. 2017 Mutation of the
07420528.2013.820739) pone.0003055) human circadian clock gene CRY1 in familial delayed
sleep phase disorder. Cell 169, 203– 215. (doi:10. 31. Ermis U, Krakow K, Voss U. 2010 Arousal thresholds mammalian sleep. Neurosci. Biobehav. Rev. 32, 8
1016/j.cell.2017.03.027) during human tonic and phasic REM sleep. J. Sleep Res. 1451– 1484. (doi:10.1016/j.neubiorev.2008.05.023)
rspb.royalsocietypublishing.org
17. Reid KJ, Zee PC. 2011 Circadian disorders of the 19, 400–406. (doi:10.1111/j.1365-2869.2010.00831.x) 47. Ohayon MM, Carskadon MA, Guilleminault C,
sleep– wake cycle. In Principles and practice of 32. R Core Team. 2016 R: a language and environment Vitiello M. 2004 Meta-analysis of quantitative sleep
sleep medicine (eds MH Kryger, T Roth, WC for statistical computing. Vienna, Austria. parameters from childhood to old age in healthy
Dement), pp. 470 –482. St. Louis, MO: Elsevier 33. Shannon W, Li T, Xian H, Wang J, Deych E, Gonzalez individuals: developing normative sleep values
Saunders. C. 2015 Functional Actigraphy Data Analysis. across the human lifespan. Sleep 27, 1255–1273.
18. Nunn CL, Samson DR, Krystal A. 2016 Shining Package ‘Actigraphy’. See CRAN repository: https:// 48. Kuna ST, Maislin G, Pack FM, Staley B, Hachadoorian R,
evolutionary light on human sleep and sleep cran.r-project.org/web/packages. Coccaro EF, Pack AI. 2012 Heritability of performance
disorders. Evol. Med. Public Health 1, 227– 243. 34. Van Someren EJ, Swaab DF, Colenda CC, Cohen W, deficit accumulation during acute sleep deprivation in
(doi:10.1093/emph/eow018) McCall WV, Rosenquist PB. 1999 Bright light therapy: twins. Sleep 35, 1223–1233. (doi:10.5665/sleep.2074)
19. Marlowe FW. 2010 The Hadza: hunter –gatherers of improved sensitivity to its effects on rest–activity 49. Ayala FJ, Campbell CA. 1974 Frequency-dependent